Part I - Analyzing Ford GoBike System Data

by Somood Musleh

Introduction

Ford GoBike System Data (38 MB, CSV File)

Ford GoBike is a regional public bicycle sharing system in the San Francisco Bay Area, California. Beginning operation in August 2013 as Bay Area Bike Share, the Ford GoBike system currently has over 2,600 bicycles in 262 stations across San Francisco, East Bay and San Jose. On June 28, 2017, the system officially launched as Ford GoBike in a partnership with Ford Motor Company. Ford GoBike, like other bike share systems, consists of a fleet of specially designed, sturdy and durable bikes that are locked into a network of docking stations throughout the city. The bikes can be unlocked from one station and returned to any other station in the system, making them ideal for one-way trips. The bikes are available for use 24 hours/day, 7 days/week, 365 days/year and riders have access to all bikes in the network when they become a member or purchase a pass.

The Data Each trip is anonymized and includes:

• Trip Duration (seconds)
• Start Time and Date
• End Time and Date
• Start Station ID
• Start Station Name
• Start Station Latitude
• Start Station Longitude
• End Station ID
• End Station Name
• End Station Latitude
• End Station Longitude
• Bike ID
• User Type (Subscriber or Customer – “Subscriber” = Member or “Customer” = Casual)

Preliminary Wrangling

	duration_sec	start_time	end_time	start_station_id	start_station_name	start_station_latitude	start_station_longitude	end_station_id	end_station_name	end_station_latitude	end_station_longitude	bike_id	user_type	member_birth_year	member_gender	bike_share_for_all_trip
0	52185	2019-02-28 17:32:10.1450	2019-03-01 08:01:55.9750	21.0	Montgomery St BART Station (Market St at 2nd St)	37.789625	-122.400811	13.0	Commercial St at Montgomery St	37.794231	-122.402923	4902	Customer	1984.0	Male	No
1	42521	2019-02-28 18:53:21.7890	2019-03-01 06:42:03.0560	23.0	The Embarcadero at Steuart St	37.791464	-122.391034	81.0	Berry St at 4th St	37.775880	-122.393170	2535	Customer	NaN	NaN	No
2	61854	2019-02-28 12:13:13.2180	2019-03-01 05:24:08.1460	86.0	Market St at Dolores St	37.769305	-122.426826	3.0	Powell St BART Station (Market St at 4th St)	37.786375	-122.404904	5905	Customer	1972.0	Male	No
3	36490	2019-02-28 17:54:26.0100	2019-03-01 04:02:36.8420	375.0	Grove St at Masonic Ave	37.774836	-122.446546	70.0	Central Ave at Fell St	37.773311	-122.444293	6638	Subscriber	1989.0	Other	No
4	1585	2019-02-28 23:54:18.5490	2019-03-01 00:20:44.0740	7.0	Frank H Ogawa Plaza	37.804562	-122.271738	222.0	10th Ave at E 15th St	37.792714	-122.248780	4898	Subscriber	1974.0	Male	Yes

(183412, 16)

<class 'pandas.core.frame.DataFrame'>
RangeIndex: 183412 entries, 0 to 183411
Data columns (total 16 columns):
 #   Column                   Non-Null Count   Dtype  
---  ------                   --------------   -----  
 0   duration_sec             183412 non-null  int64  
 1   start_time               183412 non-null  object 
 2   end_time                 183412 non-null  object 
 3   start_station_id         183215 non-null  float64
 4   start_station_name       183215 non-null  object 
 5   start_station_latitude   183412 non-null  float64
 6   start_station_longitude  183412 non-null  float64
 7   end_station_id           183215 non-null  float64
 8   end_station_name         183215 non-null  object 
 9   end_station_latitude     183412 non-null  float64
 10  end_station_longitude    183412 non-null  float64
 11  bike_id                  183412 non-null  int64  
 12  user_type                183412 non-null  object 
 13  member_birth_year        175147 non-null  float64
 14  member_gender            175147 non-null  object 
 15  bike_share_for_all_trip  183412 non-null  object 
dtypes: float64(7), int64(2), object(7)
memory usage: 22.4+ MB

	duration_sec	start_station_id	start_station_latitude	start_station_longitude	end_station_id	end_station_latitude	end_station_longitude	bike_id	member_birth_year
count	183412.000000	183215.000000	183412.000000	183412.000000	183215.000000	183412.000000	183412.000000	183412.000000	175147.000000
mean	726.078435	138.590427	37.771223	-122.352664	136.249123	37.771427	-122.352250	4472.906375	1984.806437
std	1794.389780	111.778864	0.099581	0.117097	111.515131	0.099490	0.116673	1664.383394	10.116689
min	61.000000	3.000000	37.317298	-122.453704	3.000000	37.317298	-122.453704	11.000000	1878.000000
25%	325.000000	47.000000	37.770083	-122.412408	44.000000	37.770407	-122.411726	3777.000000	1980.000000
50%	514.000000	104.000000	37.780760	-122.398285	100.000000	37.781010	-122.398279	4958.000000	1987.000000
75%	796.000000	239.000000	37.797280	-122.286533	235.000000	37.797320	-122.288045	5502.000000	1992.000000
max	85444.000000	398.000000	37.880222	-121.874119	398.000000	37.880222	-121.874119	6645.000000	2001.000000

duration_sec                 int64
start_time                  object
end_time                    object
start_station_id           float64
start_station_name          object
start_station_latitude     float64
start_station_longitude    float64
end_station_id             float64
end_station_name            object
end_station_latitude       float64
end_station_longitude      float64
bike_id                      int64
user_type                   object
member_birth_year          float64
member_gender               object
bike_share_for_all_trip     object
dtype: object

duration_sec                  0
start_time                    0
end_time                      0
start_station_id            197
start_station_name          197
start_station_latitude        0
start_station_longitude       0
end_station_id              197
end_station_name            197
end_station_latitude          0
end_station_longitude         0
bike_id                       0
user_type                     0
member_birth_year          8265
member_gender              8265
bike_share_for_all_trip       0
dtype: int64

np.int64(0)

What we will do for this dataframe is :-

• change columns dtype properly based on it's value (start_time and end_time from object to datetime , bike_share_for_all_trip from object to boolean, member_birth_year from float to int , member_gender and user_type from object to category)
• drop null row in column( start_station_id, start_station_name, end_station_id, end_station_name, member_birth_year, member_gender) .
• As always, we need to copy our dataframe before do any cleaning process, so we can refer back to the old ones.

(174952, 16)

<class 'pandas.core.frame.DataFrame'>
Index: 174952 entries, 0 to 183411
Data columns (total 16 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   duration_sec             174952 non-null  int64         
 1   start_time               174952 non-null  datetime64[ns]
 2   end_time                 174952 non-null  datetime64[ns]
 3   start_station_id         174952 non-null  float64       
 4   start_station_name       174952 non-null  object        
 5   start_station_latitude   174952 non-null  float64       
 6   start_station_longitude  174952 non-null  float64       
 7   end_station_id           174952 non-null  float64       
 8   end_station_name         174952 non-null  object        
 9   end_station_latitude     174952 non-null  float64       
 10  end_station_longitude    174952 non-null  float64       
 11  bike_id                  174952 non-null  int64         
 12  user_type                174952 non-null  object        
 13  member_birth_year        174952 non-null  float64       
 14  member_gender            174952 non-null  object        
 15  bike_share_for_all_trip  174952 non-null  bool          
dtypes: bool(1), datetime64[ns](2), float64(7), int64(2), object(4)
memory usage: 21.5+ MB

duration_sec               0
start_time                 0
end_time                   0
start_station_id           0
start_station_name         0
start_station_latitude     0
start_station_longitude    0
end_station_id             0
end_station_name           0
end_station_latitude       0
end_station_longitude      0
bike_id                    0
user_type                  0
member_birth_year          0
member_gender              0
bike_share_for_all_trip    0
dtype: int64

	duration_sec	start_time	end_time	start_station_id	start_station_name	start_station_latitude	start_station_longitude	end_station_id	end_station_name	end_station_latitude	end_station_longitude	bike_id	user_type	member_birth_year	member_gender	bike_share_for_all_trip	start_time_dayofweek	start_time_hour	member_age
0	52185	2019-02-28 17:32:10.145	2019-03-01 08:01:55.975	21.0	Montgomery St BART Station (Market St at 2nd St)	37.789625	-122.400811	13.0	Commercial St at Montgomery St	37.794231	-122.402923	4902	Customer	1984	Male	False	Thu	17	35
2	61854	2019-02-28 12:13:13.218	2019-03-01 05:24:08.146	86.0	Market St at Dolores St	37.769305	-122.426826	3.0	Powell St BART Station (Market St at 4th St)	37.786375	-122.404904	5905	Customer	1972	Male	False	Thu	12	47
3	36490	2019-02-28 17:54:26.010	2019-03-01 04:02:36.842	375.0	Grove St at Masonic Ave	37.774836	-122.446546	70.0	Central Ave at Fell St	37.773311	-122.444293	6638	Subscriber	1989	Other	False	Thu	17	30
4	1585	2019-02-28 23:54:18.549	2019-03-01 00:20:44.074	7.0	Frank H Ogawa Plaza	37.804562	-122.271738	222.0	10th Ave at E 15th St	37.792714	-122.248780	4898	Subscriber	1974	Male	True	Thu	23	45
5	1793	2019-02-28 23:49:58.632	2019-03-01 00:19:51.760	93.0	4th St at Mission Bay Blvd S	37.770407	-122.391198	323.0	Broadway at Kearny	37.798014	-122.405950	5200	Subscriber	1959	Male	False	Thu	23	60

<class 'pandas.core.frame.DataFrame'>
Index: 174952 entries, 0 to 183411
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   duration_sec             174952 non-null  int64         
 1   start_time               174952 non-null  datetime64[ns]
 2   end_time                 174952 non-null  datetime64[ns]
 3   start_station_id         174952 non-null  float64       
 4   start_station_name       174952 non-null  object        
 5   start_station_latitude   174952 non-null  float64       
 6   start_station_longitude  174952 non-null  float64       
 7   end_station_id           174952 non-null  float64       
 8   end_station_name         174952 non-null  object        
 9   end_station_latitude     174952 non-null  float64       
 10  end_station_longitude    174952 non-null  float64       
 11  bike_id                  174952 non-null  int64         
 12  user_type                174952 non-null  category      
 13  member_birth_year        174952 non-null  int64         
 14  member_gender            174952 non-null  category      
 15  bike_share_for_all_trip  174952 non-null  bool          
 16  start_time_dayofweek     174952 non-null  category      
 17  start_time_hour          174952 non-null  category      
 18  member_age               174952 non-null  int64         
dtypes: bool(1), category(4), datetime64[ns](2), float64(6), int64(4), object(2)
memory usage: 20.9+ MB

What is the structure of your dataset?

** The dataset has 174952 bike rides that happen in 2019 in the San Francisco Bay Area. The dataset contains 19 features about:**

• trip duration: how long the trip took in seconds
• start time/end time: start/end date/time for the bike rides.
• stations: start/end station, name, geolocation (latitude/longitude)
• anonymized customer data: gender, birth date and user type
• rented bikes: bike id

** The dataset was further enhanced with features that I may find neccessary to perform interesting analysis: **

• rental time: hour of the day, weekday
• customer: age

Most variables are 10 numerical, and others are 2 datetime, 2 object type , 4 category dtype and 1 is boolean type.

What is/are the main feature(s) of interest in your dataset?

I'm most interested in figuring out how trip duration is dependent on other features such as: age, start station, end station, user type, and gender from the dataset.

What features in the dataset do you think will help support your investigation into your feature(s) of interest?

I expect that trip duration will have the strongest effect on each the start stations and end stations becouse the crowded places should receive more rides. I also think that user_type, age, and gender will have effects on trip duration.

Univariate Exploration

I'll start by looking at the distribution of the main variable of interest: duration_sec.

It's seems that most data are below 2000. Let's find out the distribution.

count    174952.000000
mean        704.002744
std        1642.204905
min          61.000000
25%         323.000000
50%         510.000000
75%         789.000000
max       84548.000000
Name: duration_sec, dtype: float64

The data is not distributed normally. Let's calculate the IQR and Upper Wisker Bound :

1488.0
376.0

C:\Users\acer\AppData\Local\Temp\ipykernel_11136\2902543463.py:12: UserWarning: Attempt to set non-positive xlim on a log-scaled axis will be ignored.
  plt.axis([0, 10000, 0, 7000])

The distribution now looks closer to normal distribution. But, there is a long tail still. Let's remove the long tail.

<class 'pandas.core.frame.DataFrame'>
Index: 174068 entries, 4 to 183411
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   duration_sec             174068 non-null  int64         
 1   start_time               174068 non-null  datetime64[ns]
 2   end_time                 174068 non-null  datetime64[ns]
 3   start_station_id         174068 non-null  float64       
 4   start_station_name       174068 non-null  object        
 5   start_station_latitude   174068 non-null  float64       
 6   start_station_longitude  174068 non-null  float64       
 7   end_station_id           174068 non-null  float64       
 8   end_station_name         174068 non-null  object        
 9   end_station_latitude     174068 non-null  float64       
 10  end_station_longitude    174068 non-null  float64       
 11  bike_id                  174068 non-null  int64         
 12  user_type                174068 non-null  category      
 13  member_birth_year        174068 non-null  int64         
 14  member_gender            174068 non-null  category      
 15  bike_share_for_all_trip  174068 non-null  bool          
 16  start_time_dayofweek     174068 non-null  category      
 17  start_time_hour          174068 non-null  category      
 18  member_age               174068 non-null  int64         
dtypes: bool(1), category(4), datetime64[ns](2), float64(6), int64(4), object(2)
memory usage: 20.8+ MB

C:\Users\acer\AppData\Local\Temp\ipykernel_11136\2902543463.py:12: UserWarning: Attempt to set non-positive xlim on a log-scaled axis will be ignored.
  plt.axis([0, 10000, 0, 7000])

Trip duration is mostly centered on values which are less than 4000 seconds with peak around 550 seconds. Number of Trips values first increases starting from around 5000 values to 6500 values at around 600 seconds but then starts to fall below at 1000 values under 2000 seconds.

Let's look at member age now.

count    174068.000000
mean         34.191810
std          10.117383
min          18.000000
25%          27.000000
50%          32.000000
75%          39.000000
max         141.000000
Name: member_age, dtype: float64

Note: max age is 141! Not usual.

The data is not distributed evenly. Let's calculate the IQR and Upper Wisker Bound

np.float64(57.0)

<class 'pandas.core.frame.DataFrame'>
Index: 170557 entries, 4 to 183411
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   duration_sec             170557 non-null  int64         
 1   start_time               170557 non-null  datetime64[ns]
 2   end_time                 170557 non-null  datetime64[ns]
 3   start_station_id         170557 non-null  float64       
 4   start_station_name       170557 non-null  object        
 5   start_station_latitude   170557 non-null  float64       
 6   start_station_longitude  170557 non-null  float64       
 7   end_station_id           170557 non-null  float64       
 8   end_station_name         170557 non-null  object        
 9   end_station_latitude     170557 non-null  float64       
 10  end_station_longitude    170557 non-null  float64       
 11  bike_id                  170557 non-null  int64         
 12  user_type                170557 non-null  category      
 13  member_birth_year        170557 non-null  int64         
 14  member_gender            170557 non-null  category      
 15  bike_share_for_all_trip  170557 non-null  bool          
 16  start_time_dayofweek     170557 non-null  category      
 17  start_time_hour          170557 non-null  category      
 18  member_age               170557 non-null  int64         
dtypes: bool(1), category(4), datetime64[ns](2), float64(6), int64(4), object(2)
memory usage: 20.3+ MB

The data is not distributed evenly. Many cyclists are around 30 years old. Not many cyclists are over 40 years old.

8:00 AM and 5:00 PM are the two peak hours with the most cyclists. And Most trips are during the day not in night.

There are about three times as many male cyclists as female cyclists.

There are more subscribers than customer.

Discuss the distribution(s) of your variable(s) of interest. Were there any unusual points? Did you need to perform any transformations?

The main variable of interest, duration_sec, had a highly skewed distribution. Most trips were shorter than 2,000 seconds, with a peak around 550 seconds. There was a long tail in the distribution due to a small number of trips with excessively high durations, which made it challenging to interpret patterns.

To address this, a log transformation was applied to the duration_sec variable to make the distribution closer to normal. Despite the transformation, some long-tail effects remained, prompting further data cleaning to exclude trips exceeding 4,000 seconds. This threshold was chosen based on the interquartile range (IQR) and upper whisker bound, which showed that these outliers were likely unrepresentative of typical trips.

Other notable unusual points included:

• member_age: The distribution showed most users were around 30 years old, with a sharp decline after 40. There was an extreme outlier with a calculated age of 141, which was likely due to erroneous or missing data. Data cleaning was performed to exclude users older than 60 years. This was based on domain knowledge that such values were implausible for regular cyclists.
• start_time_hour: The distribution revealed two prominent peaks at 8:00 AM and 5:00 PM, suggesting heavy usage during typical commute hours. This pattern aligns with the expected behavior for a bike-sharing system in an urban area.

Of the features you investigated, were there any unusual distributions? Did you perform any operations on the data to tidy, adjust, or change the form of the data? If so, why did you do this?

Several operations were performed to tidy and clean the data:

1. Null values: Rows with missing data in key columns like start_station_name, end_station_name, member_birth_year, and member_gender were dropped to ensure the dataset's integrity.
2. Data types: Variables like start_time and end_time were converted to datetime formats, while categorical variables such as member_gender were adjusted to improve processing.
3. Derived features: Additional variables, such as start_time_dayofweek, start_time_hour, and member_age, were created to enable more meaningful analysis.
4. Outlier removal: Long-duration trips and implausibly high ages were removed to focus on typical user behavior and reduce skewness in the distributions.

These transformations improved the data's usability and ensured the insights were accurate and reflective of real-world patterns.

Bivariate Exploration

<class 'pandas.core.frame.DataFrame'>
Index: 170557 entries, 4 to 183411
Data columns (total 19 columns):
 #   Column                   Non-Null Count   Dtype         
---  ------                   --------------   -----         
 0   duration_sec             170557 non-null  int64         
 1   start_time               170557 non-null  datetime64[ns]
 2   end_time                 170557 non-null  datetime64[ns]
 3   start_station_id         170557 non-null  float64       
 4   start_station_name       170557 non-null  object        
 5   start_station_latitude   170557 non-null  float64       
 6   start_station_longitude  170557 non-null  float64       
 7   end_station_id           170557 non-null  float64       
 8   end_station_name         170557 non-null  object        
 9   end_station_latitude     170557 non-null  float64       
 10  end_station_longitude    170557 non-null  float64       
 11  bike_id                  170557 non-null  int64         
 12  user_type                170557 non-null  category      
 13  member_birth_year        170557 non-null  int64         
 14  member_gender            170557 non-null  category      
 15  bike_share_for_all_trip  170557 non-null  bool          
 16  start_time_dayofweek     170557 non-null  category      
 17  start_time_hour          170557 non-null  category      
 18  member_age               170557 non-null  int64         
dtypes: bool(1), category(4), datetime64[ns](2), float64(6), int64(4), object(2)
memory usage: 20.3+ MB

To start off with, I want to look at the pairwise correlations present between features in the data.

Pairwise Correlation Heatmap

• Explore relationships between numeric variables.
• A heatmap shows the correlation coefficients, where:
  • 1.0 indicates a strong positive correlation.
  • -1.0 indicates a strong negative correlation.
  • 0 indicates no correlation.
• Use a diverging colormap to highlight patterns.

The correlation cofficients indicate there might be correlations between the following pairs of variables: (start_station_id, end_station_id), (start_station_latitude, start_station_longitude), and (end_station_latitude, end_station_longitude). However, there are no strong evidences to support these. So these might happen just by accident.

Pair Plots for Numeric Variables

• Use pair plots to explore relationships between specific numeric variables.
• Sample data for better clarity and faster rendering.
• Diagonal: Histograms of individual variables.
• Off-diagonal: Scatter plots of variable relationships.

df_clean.shape= (170557, 19)
df_clean_samp.shape= (500, 19)

<seaborn.axisgrid.PairGrid at 0x1ec56cf8380>

There is no specific relationship (like linear) between duration_sec and member_age.

Boxplots: Categorical vs. Numeric Variables

• Visualize the relationships between categorical and numeric variables.
• Use a sample to reduce clutter in the plots.
• Boxplots show the spread of numeric variables (e.g., trip duration) for each category.

<Figure size 1500x1500 with 0 Axes>

Observations from above plots:

• Bike rides on Saturday and Sunday have longer durations compared to bike rides on other weekdays.
• Bike rides start between 5:00 AM to 6:00 PM have the longest mean durations compared to other time frames.
• On average, female cyclists ride longer than male cyclists.
• On average, cyclists who are not members (or suscribers) ride longer than cyclists who are subscribers.
• The average age of cyclists on Tuesday and Wednesday are older than that of cyclists on other days.
• The average age of cyclists on Sunday is younger than that of cyclists on other days.
• The average age of cyclists who start cycling between 6:00 AM to 7:00 AM is older than that of cyclists who start cycling at other time frames.
• The average age of male cyclists is older than that of female cyclists.
• The average age of cyclists who are subscribers is older than that of cyclists who are not subscribers.
• Interestingly, we can see that higher percentage of customer rides longer trips then subscriber although number of subscriber riders are very high then customer.

Countplots for Categorical Relationships

• Compare counts of categorical variables across different groupings.
• For example:
  • How does trip hour relate to user type?
  • How does member gender compare with trip duration?
• Each subplot focuses on a different categorical comparison.

Observations from above plots:

• There are the most subscriber type cyclists at 5:00 PM. There are the most non-subscriber type cyslists at 5:00 PM too.
• There are the most male cyclists at 5:00 PM. There are the most female cyclists at 8:00 AM and 5:00 PM.
• There are the most cyclists at 5:00 PM on Tuesday.
• There are the most male cyclists on Tuesday. There are the most female cyclists on Tuesday too.
• There are the most subscriber type cyclists on Tuesday. There are the most customer type cyclists on Saturday.
• Most subscriber type cyclists are male. Most non-subscriber type cyclists are male too.

Scatter Plots: Member Age vs. Trip Duration

• Explore how trip duration varies with member age.
• Use a log scale for better visualization of large values.
• Randomly sample data points for clarity.

Although we are still not seeing linear relationship between duration_sec and member_age, the data looks distribute more even compared to the scatter plot drawn eariler.

By looking at the chart above, we can see that most frequent users aged between 20 and 45. As remark, duration is registered by younger members.

Observations:

• The shapes of the violins in the plots on the left side (with 'log_duration_sec' as y-axis) are more even compared to the shapes of the violins in the plots on the right side (with member_age as y-axis). The transformation make the violins on the left side looks more even.
• Look at the plot on 1st row, 1st column: there are more bike rides with durations close to mean duration time (10 minutes) on the weekdays compared to the bike rides on the weekends.
• Look at the plot on 3rd row, 1st column: the mean bike duration for female cyclists is longer than the the mean duration of the male cyclists.
• Look at the plot on 4th row, 1st column: the mean bike duration for customer type cyclists is longer than the the mean duration of the subscriber type cyclists.

Talk about some of the relationships you observed in this part of the investigation. How did the feature(s) of interest vary with other features in the dataset?

Relationships with Features of Interest

1. Trip Duration (duration_sec) vs. Member Age (member_age):
   • The scatter plot revealed that younger riders (ages 20–40) tend to have shorter trip durations compared to older age groups.
   • A log transformation on trip duration clarified patterns and outliers, as some trips had unusually long durations.

2.Trip Start Time (start_time_hour) vs. User Type:

  - Peak travel times (8–9 AM and 5–6 PM) were dominated by subscribers, suggesting their usage is aligned with commuting patterns.
  - Casual users had a more even distribution of start times throughout the day, likely tied to recreational use.

3. Trip Start Time (start_time_hour) vs. Member Gender:
   • Male riders were the majority across most hours.
   • Female riders followed a similar pattern but were less represented overall.

Did you observe any interesting relationships between the other features (not the main feature(s) of interest)?

Other Interesting Relationships

1. Member Age Distribution:
   • Members aged 30–40 constituted the majority, with a gradual decline in the number of riders as age increased.
   • Few riders were under 20 or above 60, indicating that the service primarily caters to middle-aged individuals.

2. User Type (user_type) vs. Gender:
   • Subscriber-type users were predominantly male.
   • Female riders were more likely to be casual users than subscribers, compared to their male counterparts.

3. Station-Based Trends:
   • Certain stations had higher trip frequencies, particularly in areas near central business districts or popular recreational zones.
   • Starting and ending station IDs showed a correlation, indicating many trips began and ended at the same location (likely for recreational loops).

4. Day of Week Trends:
   • Weekdays had a higher volume of trips by subscribers, consistent with commuting patterns.
   • Weekends saw a rise in casual user activity, likely driven by leisure trips.

Insights

• The dataset suggests that bike-sharing services primarily serve urban commuters and recreational riders, with distinct patterns across time, age, and user type.
• Outliers in trip duration and user behavior indicate opportunities for targeted analysis (e.g., identifying unusually long trips or underutilized stations).

Multivariate Exploration

The main thing I want to explore in this part of the analysis is how the three categorical measures of user type and gender into the relationship between trip duration and age.

• Visualization: A grid of 2D histograms showing trip duration vs. member age for each day of the week.
• Observations:
  • Younger riders (ages 20–40) consistently have shorter trip durations across all days.
  • Weekends show a broader distribution of trip durations, possibly reflecting more recreational usage.
  • Trips on weekdays are concentrated in specific duration ranges, aligning with commuter behavior.

• Visualization: A grid of 2D histograms showing trip duration vs. member age for each hour of the day.
• Observations:
  • Peak commuter hours (8–9 AM and 5–6 PM) have a higher density of shorter trips by younger riders.
  • Non-peak hours show a wider spread in trip durations, particularly among older age groups.
  • Riders above 50 years tend to have fewer trips during peak hours but longer trip durations overall.

• Visualization: 2D histograms showing trip duration vs. member age for each gender.
• Observations:
  • Male riders dominate across all age groups and duration ranges.
  • Female riders have fewer trips overall but exhibit similar patterns, with shorter trips being more common.
  • Other/unspecified genders have limited data, resulting in sparser distributions.

• Visualization: 2D histograms showing trip duration vs. member age for subscribers and casual users.
• Observations:
  • Subscribers have shorter trips concentrated among younger members, indicative of regular commuting behavior.
  • Casual users show a wider distribution of trip durations and ages, suggesting a mix of recreational and occasional usage.

• Visualization: A detailed grid of heatmaps showing trip duration vs. member age for each combination of day and hour.
• Observations:
  • Morning and evening peaks on weekdays show dense clusters of shorter trips by younger riders.
  • Midday on weekends sees a broader spread of trip durations and ages, reflecting recreational activity.
  • Late-night trips are sparse but show higher variability in duration.

These visualizations highlight distinct patterns in trip duration and member age influenced by factors such as time of day, day of the week, gender, and user type. Subscribers exhibit more structured behavior, while casual users have diverse trip patterns.

Observation:

• Male riders dominate across all age groups and duration ranges.
• Female riders have shorter trip durations overall, with fewer long trips.
• Other/unspecified genders have limited data points, showing less clear trends.

Observation:

• Male and female riders share similar patterns of short trip durations.
• Variability in trip duration increases slightly for older age groups across genders.
• Other gender category data is sparse and does not show clear patterns.

Observation:

• Subscribers tend to have shorter trip durations across all age groups.
• Customers have more variability in trip durations, particularly at younger ages.
• The majority of trips by both user types are under 5,000 seconds.

Observation:

• Subscribers show dense clusters of shorter trips, especially for younger riders.
• Customers exhibit wider variability in trip durations, likely reflecting recreational usage.
• Across user types, shorter trips dominate, particularly for riders aged 20–40.

Let's devide bikers into 2 groups by age: one group has cyclists below 40 years old, another group has cyclists above 40 years old.

Observations:

• By looking at the errorbars in both pointplots, we can tell that there are more younger cyclists (age < 40) across 7 days and 24 hours.
• Cyclists on Saturdays and Sundays ride longer than cyclists on other days.

Observations:

• By looking at the errorbars (first plot, other cyclists; second plot, female cyclists and other cyclists) in both pointplots, we can tell that there are less cyclists who are older than 40 years old.
• Female cyclists ride longer than male cyclists.

Observations:

• By looking at the errorbars in both pointplots, we can tell that there are less cyclists who are older than 40 years old.
• Customer type cyclists ride longer than subscriber type cyclists.

Talk about some of the relationships you observed in this part of the investigation. Were there features that strengthened each other in terms of looking at your feature(s) of interest?

Relationships Observed

• Age and Trip Duration: Younger riders (ages 20–40) consistently recorded shorter trip durations, particularly on weekdays. This aligns with regular commuting behavior.
• Day of the Week and Trip Duration: Weekends exhibited longer and more varied trip durations compared to weekdays, suggesting recreational or leisure-oriented usage.
• User Type and Trip Duration: Subscribers predominantly had shorter trip durations, indicating habitual, utility-driven trips. In contrast, casual users showed a broader distribution, reflective of occasional or exploratory use.
• Gender and Trip Duration: Male riders dominated across all categories. Female riders showed slightly longer trip durations on average, particularly on weekends.

Strengthened Relationships

• User Type and Age: Subscribers under 40 exhibited concentrated patterns of short trips, reinforcing the idea of routine commuting.
• Day of the Week and User Type: On weekdays, subscribers showed highly structured patterns of shorter trips, while casual users exhibited greater variability, especially on weekends.
• Gender and Day of the Week: Female riders appeared to have slightly longer trips on weekends, particularly in the recreational time frames.

Were there any interesting or surprising interactions between features?

Surprising Interactions

• Peak Hours and Age: While younger riders had higher densities of shorter trips during peak hours, older riders had longer trips overall, even during non-peak hours. This suggests a difference in trip purpose between age groups.
• Gender and Duration: Although male riders outnumbered others, female riders often recorded longer average trip durations, particularly for older age groups. This was unexpected given the smaller sample size for female riders.
• Casual Users on Weekdays: Some casual users exhibited trip durations similar to subscribers on weekdays, which might indicate overlap in usage patterns, such as tourists commuting or new users trying the service.

Reflection

These interactions reveal nuanced usage patterns among different demographics and user types. They highlight how specific features (e.g., age, user type, and time) work together to shape trip characteristics.

Conclusions

Summary of Main Findings

1. Trip Duration and Age:

   • Younger riders (20–40 years) took shorter trips, with durations decreasing further during peak commuter hours.
   • Older riders (40+ years) tended to have longer trips, particularly during non-peak hours.

2. Day of the Week and Trip Purpose:

   • Weekdays showed concentrated, shorter trips consistent with commuting patterns.
   • Weekends exhibited broader trip durations, likely reflecting recreational use.

3. User Types:

   • Subscribers predominantly had shorter, structured trips, aligning with routine commuting.
   • Casual users showed a diverse range of trip durations and ages, suggesting more recreational or occasional use.

4. Gender Differences:

   • Male riders dominated across all age and duration categories.
   • Female riders took fewer trips but often recorded longer average durations, especially on weekends.
   • Other/unspecified genders showed sparse data, limiting insights.

5. Time of Day Patterns:

   • Morning and evening peak hours were dominated by younger riders taking short trips.
   • Midday hours, especially on weekends, saw broader distributions of ages and trip durations.

6. Interactive Insights:

   • Age, user type, and trip duration revealed clear behavioral distinctions, with structured, shorter trips by younger subscribers versus longer, variable trips by casual users.

---

Reflection on Data Exploration Steps

• Data Cleaning:

  • Successfully handled missing and inconsistent values (e.g., member gender, trip duration outliers).
  • Introduced computed features such as member age and log-transformed trip durations for better analysis.

• Univariate Exploration:

  • Observed central tendencies and distribution of individual features like age, trip duration, and user types.

• Bivariate Exploration:

  • Identified correlations between trip duration and factors like age, gender, and user type.

• Multivariate Exploration:

  • Revealed complex interactions involving time of day, day of the week, and demographic variables.
  • Heatmaps and scatterplots facilitated insights into rider behavior across categories.

This exploration uncovered key behavioral patterns among riders, helping to understand how demographic, temporal, and categorical factors shape trip characteristics.

[NbConvertApp] Converting notebook Part_I_exploration_template_(1).ipynb to html
[NbConvertApp] WARNING | Alternative text is missing on 34 image(s).
[NbConvertApp] Writing 4727950 bytes to Part_I_exploration_template.html